The digitising magic is done by a chip - called an "Analog to Digital Converter" or "A->D" chip (it measures the signal's voltage the appropriate number of times per second and spits out a digital signal). That chip needs to connect to other components (and connectors), so it is put onto a circuit board. A circuit board can be designed to fit into a box. Or the board, itself, can be a pci/isa card. Your typical box will also have hardware to convert the signal into mpeg - which takes away most of your options for processing the capture. A box's mpeg solution will be optimised for real-time speed, at the sacrifice of quality, so it will suck. When you use a capture card, its best to capture to Huffyuv, which is lossless. Then you end up with a gigantic file (80 Gigs per half-hour) that you can process to perfection.
Avisynth and Virtualdub have various methods for deinterlacing & Inverse Telecine. Your camcorder footage is probably 60 fields per second, so that leaves out Inverse Telecine (which is for progressive footage). Deinterlacers can do a fair job of blending or interpolating fields to get progressive. But since there's a lot of action in your footage, its probably better to leave it interlaced, unless you want to make it 60 frames per second, for the computer monitor or an advanced TV (standard NTSC tvs are 30 frames per second = 60 fields per second). I can't recommend plugins, I've only done IVTC (Inverse Telecine). There are forum threads (Doom9) for each plugin, and threads about general recomendations (Videohelp), where experienced people will talk over your head and tell you to read the often-ambiguous & sparse & often-undefined documentation, and tell you to experiment and learn.
Not that I'm ranting or anything.